Thursday, 4 December, 2025

14:00-14:40 Albert Diaz-Guilera (Universitat de Barcelona), Synchronization and Beyond: Exploring Dynamics in Complex Networks

TBA

14:45-15:10 Riccardo Muolo (RIKEN & Institute of Science Tokyo), Higher-order phase reduction through parametrization

Recent studies have shown that non-pairwise interactions play a crucial role in shaping synchronization dynamics.

Such interactions can have different origins.  For instance, in systems with physical states $x_k, x_j, x_l \in \mathbb{R}$, a non-pairwise coupling may take the form $G(x_k, x_j, x_l) = x_kx_jx_l$, whereas in phase models with $\theta_k, \theta_j, \theta_l \in \mathbb{T}$, it may appear as $g(\theta_k, \theta_j, \theta_l) = \sin(\theta_k+\theta_j-2\theta_l)$.

The link between such physical and phase-level interactions is established through phase reduction for weakly coupled oscillators, generally expressed as an asymptotic expansion in the coupling strength~$\varepsilon$.

Classical first-order phase reductions, such as the Kuramoto one, capture only pairwise interactions, but higher-order reductions, such as the León and Pazó one, reveal the emergence of non-pairwise terms.  These can arise either directly from nonpairwise couplings in the physical variables or at higher orders even when the physical coupling is pairwise.

Here, we employ a parametrization-based method for higher-order phase reduction that captures both routes simultaneously by approximating an invariant torus in the full system phase space, providing a unified description of nonpairwise synchronization dynamics.

15:15-15:30 Mircheski Petar (Institute of Science Tokyo), Control of chimera states

TBA

15:45-16:10 Praful Gagrani (University of Tokyo), Large deviation theory of growing chemical reaction networks

TBA

16:15-16:40 Barbieri Ettore (JAMSTEC), Reconstructing Ecological Food Webs from Species Lists Using Large Language Models

This study introduces a simple and scalable way to rebuild ecological food webs when direct feeding data are missing. We use Large Language Models (LLMs) to read species names and automatically assign each species to a broad trophic group, such as producer, zooplankton, benthic invertebrate, fish, bird, or marine mammal. Based on ecological knowledge, a set of feeding rules then defines which groups can eat which others, creating a predator–prey network for each ecosystem. The method was applied to 158 food webs containing different numbers of species. For each web, we generated an adjacency matrix (showing who eats whom), the trophic group of each species, and simple network measures such as the number of links and density. 

By combining the language understanding of LLMs with basic ecological logic, this approach can infer realistic food-web structures from minimal information—just a list of species. The results can be used to compare ecosystems, study patterns of energy flow, or serve as a starting point for more detailed models when observational data are scarce. This work shows how modern language models can help scientists turn unstructured biological lists into structured ecological networks in a transparent and reproducible way

16:45-17:25 Shigefumi Hata (Kagoshima University), Synchronization facilitated by frequency differences: Dynamics of coupled-oscillator systems with damaged elements

This study investigated synchronization dynamics of coupled-oscillator systems in which some of oscillators are damaged and lose their autonomous oscillations. The damaged elements are modeled using a damped oscillator; thus, the system is composed of both limit-cycle oscillators and damped oscillators. In this system, as is commonly observed in conventional coupled limit-cycle oscillators, synchronization among oscillators is destroyed when the difference between the natural frequencies of the oscillators increases. However, in the presence of damped oscillators, synchronization can be facilitated by further increasing the frequency difference from the desynchronization state. We conducted numerical simulations on coupled Stuart-Landau oscillators and investigated this reentrance of synchronization systematically. We also propose an approximate theory to predict the stability of the synchronization state based on a linear stability analysis of the fixed point, which reveals the appearance of the Hopf mode. Using this theory, we argue that the reentrance of synchronization driven by increasing frequency differences can be observed in a wide range of coupled-oscillator systems with damaged elements.

Friday, 5 December, 2025

1​0​:00​​-1​0:​5​​0 合原 一幸 (東大, remote), Early Warning Signals and Treatments for Asymptomatic Mebyo States

I review our recent studies on early warning signals and treatments of asymptomatic Mebyo states on the basis of Dynamical Network Biomarkers(DNB) to detect early warning signals of Mebyo states and network control theory to restabilize Mebyo states into healthy states as treatments.

​1​1​:​00​-​1​1:​40​ ​横山 寛 (岡山大), Estimating the dynamical and causal structures in synchronous neural oscillations using a data-driven approach

Understanding the causal mechanisms in brain network dynamics is one of the primary goals in human brain mapping studies. To achieve this aim, it is essential to interpret the mechanisms underlying the observed neural data from both experimental and computational perspectives. In this context, the presenter is engaged in a neuroscience study aimed at developing a method for analyzing observed neural data using data-driven modeling approaches. In this presentation, I would like to explain our recent works using two different approaches: (1) data-assimilation methods—which combine observational data with computational models to estimate hidden brain states—and (2) causal discovery-based methods—which are used to infer causal relationships from observed neural data—to reveal the functional mechanisms in the brain network dynamics in a data-driven manner.

The presentations will be organized into two parts. The first part is related to our recently published papers applying data assimilation techniques to reconstruct and interpret the internal brain state and its function in a data-driven manner. The second part is related to recent preliminary studies for estimating causal structures and interventional effect for synchronous neural data. In this presentation, I would like to explain both the numerical and physiological validity of our proposed method based on the results when applying synthetic and human/animal observational data.

1​1:45-12:25 寺前 順之介 (京大), Random matrix approach to the capacity-amplification trade-off in recurrent memory networks

Conventional models of sensory coding and memory retrieval have primarily focused on steady-state neural activity, represented as stable fixed points or attractors. However, recent experimental results have revealed that transient neural trajectories, occurring before convergence to steady states, are more informative for discriminating sensory stimuli than the steady-state activity. Here, we propose a theoretical framework for a recurrent network model of stimulus-dependent dynamical memories. By developing a mathematical theory of transient memory dynamics based on random matrix theory, we reveal a fundamental trade-off between the memory capacity of the network and the transient amplification of embedded trajectories. These findings imply that the brain must balance the number of possible trajectories with their separability for efficient processing of sensory information through dynamic representations.

14:00-14:25 名村 憲尚(東京科学大), Higher-order Kuramoto models from arbitrary limit-cycle oscillators

TBA

14:25-14:40 Taichi Yamamoto (University of Tokyo GSFS), Kernel EDMD for robust data-driven modeling of oscillation

Vibrational dynamic systems such as the heart, brain waves, and fluid vortices exist widely in reality. Extracting key degrees of freedom (patterns) from the time-series data obtained from these systems is essential for understanding and controlling the phenomena. Extended Dynamic Mode Decomposition (EDMD), which estimates the Koopman operator and its eigenfunctions, has been studied as a method for this purpose.

Although Koopman eigenfunctions exist in infinite numbers, it is known that only a few principal eigenfunctions represent the key degrees of freedom. However, research evaluating the accuracy of EDMD has focused on assessing the prediction accuracy obtained as the sum of all eigenmodes, without paying attention to the accuracy of a single eigenmode. Furthermore, the theory of EDMD assumes perfect observations, and its accuracy under observational noise has been little discussed.

Therefore, this study focuses on the estimation accuracy of the principal eigenmodes, and how robust the results are when observational noise is present or when the sampling interval varies. We organize the EDMD and kernel EDMD variants and apply them to Stuart-Landau equation and cylinder Kalman vortex. The results reveal that dimensionality reduction and regularization enhance both accuracy and robustness.

14:40-14:55 Yutaro Takahashi (Grad. Sch. Info., Kyoto Univ.), Self-organization of intestinal peristalsis: A phase oscillator model with time-varying natural frequencies

Peristalsis refers to the transport of digested material driven by propagating waves of contraction and relaxation in intestinal muscles. Although rhythmic coordination of the entire intestine is necessary for this transport, the essential mechanism underlying this coordination remains unclear. Developmental studies of the chicken embryo small intestine have shown that peristaltic origins (sites where peristaltic waves originate) initially appear irregularly. However, as development progresses, these origins align and become spaced at appropriate intervals. This observation suggests the existence of a self-organization mechanism for peristaltic origins within the intestine. Furthermore, another study using the chicken cecum confirmed that the contraction frequency is highest at the peristaltic origins and changes smoothly along the intestinal tract. Based on these experimental findings, this study models the intestinal tract as a one-dimensional chain of coupled phase oscillators, where each oscillator represents a segment of the intestine. We also introduce a mechanism for the natural frequency of each oscillator to evolve over time, constructing a model consistent with these experimental observations. Numerical simulations confirm that our model successfully reproduces the self-organization of peristaltic origins.

14:55-15:10 鈴木 隆人 (京大), Anomalous Enhancement of Yield Strength due to Static Friction

Many structures around us achieve mechanical stability against their own weight through friction. Examples include arch bridges, house-of-cards construction, and sandpile stabilized by their angle of repose. In these systems, friction and geometry work together to suppress sliding and enables mechanical equilibrium. But what (if any) mechanism contributes to the strength of such friction-stabilized configurations?

In this talk, we investigate a minimal model consisting of three cylindrical particles stacked via side-to-side contact under gravity, forming a triangular arrangement. A quasi-static compressive force is applied from above via a wall. As the force increases, the structure eventually collapses due to sliding at the contact with the floor. We define the threshold force required to induce this failure as the yield force, and study its dependence on the floor friction coefficient and the stiffness of the cylinders. 

Surprisingly, in the rigid-body case (i.e., no deformation), we find a sharp transition: the yield force diverges at a critical friction coefficient \mu_c ~ 0.268. To investigate more realistic conditions, we employ the discrete element method (DEM) to analyze a pile of elastic cylinders. We numerically observe a singular behavior as the dimensionless effective elastic modulus becomes large. Furthermore, we derive a unified scaling function to describe this singularity.

15:10-15:50 多賀 圭理 (東京理科大), Koopman analysis for the coupled oscillator systems via Perron-Frobenius operator

TBA

Saturday, 6 December, 2025 

​

1​0:00​-1​0​:​15 末谷 大道 (大分大), An overview of reservoir computing

TBA

​10:15-10:55 田中 剛平 (名古屋工大), Diversity-aware reservoir computing

Many real-world systems consist of interacting heterogeneous components. Inspired by this fact, we have been trying to develop computational models considering the diversity of system components, aiming to explore flexible information processing. In this talk, I will introduce relevant research examples related to reservoir computing, including the diverse-timescale echo state network (DTS-ESN) for enhancement of predictive performance on multi-timescale dynamical systems [1, 2], the simulator development on memristor-network reservoir computing systems considering the variability of memristive elements [3], and the reservoir state analysis method for time series anomaly detection [4].

[1] Tanaka et al., Physical Review Research, Vol.4, L032014 (2022)

[2] Tanaka, Springer. DOI: 10.1007/978-981-96-8383-3_2 (2026)

[3] Tanaka and Nakane, Scientific Reports, Vol.12, 9868 (2022)

[4] Tamura et al., TechRxiv, preprint. DOI: 10.36227/techrxiv.174378296.63561723/v1.

11:00-11:20 大石 悟 (阪大), Stabilize chaotic dynamics reconstruction in reservoir computer

TBA

11:20-11:35 中村 仁 (阪大), Learning Dynamics of Recurrent Parameter-Sharing Transformers via Local Lyapunov Exponents

TBA

11:35-12:15 塩澤 航太 (崇城大), Data-driven estimation of covariant Lyapunov vectors using reservoir computing

Covariant Lyapunov vectors (CLVs) have played a central role in characterizing the geometrical structure of chaotic attractors. Despite their prominent importance, estimating CLVs solely from data remains a challenging problem. Here, we propose a fully data-driven approach based on reservoir computing. We test our method through numerical experiments.

14:00-14:40 香取 勇一 (はこだて未来大), Reservoir-Based Predictive Coding for Multisensory Integration and Noise-Adaptive Neural Computation

Predictive coding describes cortical computation as the interaction between internal predictions and sensory inputs, while reservoir computing provides a nonlinear dynamical substrate requiring minimal training. This talk presents a unified Predictive Coding with Reservoir Computing (PCRC) framework in which recurrent reservoirs generate predictions and compute prediction errors for sensory signals.

Building on this architecture, I introduce a multisensory integration model that processes auditory and visual information through interconnected reservoir modules. The model reconstructs visual patterns from auditory cues and adaptively adjusts feedback gain by estimating sensory noise from prediction-error statistics and mapping it through a sigmoidal control law. This mechanism reproduces inverse effectiveness and achieves robust recognition across varying noise levels, offering a flexible computational account of multisensory processing.

14:45-15:25 堀田 武彦 (大阪公大), Large Deviation Analysis in Strong and Weak Generalized Synchronization

In a drive-response system, under certain conditions, a functional relationship between the response system and the drive system on the attractor of the whole system arises. We discuss its "smoothness" in relation to the joint large deviation property of the least finite-time Lyapunov exponents of the drive system and the largest finite-time Lyapunov exponents of the response system.

Collaborators: Katsuya Ouchi and Hiromichi Suetani

15:30-16:10 斉藤 朝輝 (はこだて未来大), True orbit computation of chaotic maps and its application to pseudorandom number generation

In this talk, we introduce a method for computing true orbits of piecewise linear and linear fractional maps with rational coefficients by using algebraic numbers of suitably chosen degree.  We also report that, by applying this method to maps such as the Bernoulli map and the continued-fraction transformation---maps for which conventional methods have difficulty producing faithful simulations---we are able to generate sufficiently long true orbits that exhibit the same properties as typical orbits of these maps.  We also explain applications of this method to pseudorandom number generation.

16:25-17:05 石崎 龍二(福岡県立大), Application of Self-Exciting Point Process Models to Multiple Volcanoes

Volcanic eruptions and related earthquakes often show clustering in time, indicating self-exciting and nonlinear interaction behavior. In this study, self-exciting point process models are applied to multiple volcanoes to examine differences in their temporal activity patterns. The model parameters are estimated for each volcano to evaluate the degree of self-excitation and triggering, providing a statistical basis for understanding nonlinear dynamics in volcanic activity.

17:10-17:50 梅野 健 (京大), Infinite Dimensional Chaotic Synchronization and a Statistical Mechanics View for Earthquake Prediction

A model for infinite dimensional chaotic synchronization is constructed with the ergodic theory, which  is found to have a phase transition between infinite dimensional synchronization-desynchronized chaos like the Bose-Einstein Condensation.   Furthermore, a statistical mechanics view (such as fluctuation-dissipation theorem for diffusion phenomena) is applied to pre-slip phenomena, a key precursor phenomena 1〜2 hours before large earthquake, which were recently detected by our group by a novel correlation analysis, and their relation with infinite dimensional chaotic synchronization is discussed.  

Sunday, 7 December, 2025 

10:00-​1​0:​50 Kunihiko Kaneko (Niels Bohr Institute & University of Tokyo, remote), 多時間スケール力学系と生命現象

生命システムを大自由度の力学系としてみたときに、非常に異なる時間スケールを含む点が特筆される。本講演ではまず、生命システムを考える上でどのように力学系研究を展開させていくべきか (Dynamical Systems ++)を議論したうえで、

＊発生過程の進化モデルにより、遅いスケールと速いスケールが分離し、自身を制御、発展させていく力学系があらわれうること

＊遅いエピジェネティック過程と速いタンパク発現ダイナミクスを結合させたモデルにより、ロバストな分化過程を示すWaddington地形が形成されること、その際に速い発現振動の再獲得により（iPS細胞で示されたような）分化の巻き戻し（reprogramming）が可能であること

＊脳神経系での速い/遅い力学系がベイズ推定を可能にすること

を紹介する予定である。

References:

Kaneko, K.(2025) Universal Biology: The Physics of Life through the Macro-Micro Consistency Principle, Cambrdge Univ. Prss

Kaneko, K. (2015). Dynamical Systems++ for a Theory of Biological System. In Chaos, Information Processing And Paradoxical Games: The Legacy Of John S Nicolis (pp. 345-354).

Kohsokabe, T., Kuratanai, S., & Kaneko, K. (2024). Developmental hourglass: Verification by numerical evolution and elucidation by dynamical-systems theory. PLOS Computational Biology, 20(2), e1011867.

Matsushita, Y., & Kaneko, K. (2020). Homeorhesis in Waddington's landscape by epigenetic feedback regulation. Physical Review Research, 2(2), 023083.

Matsushita, Y., Hatakeyama, T. S., & Kaneko, K. (2022). Dynamical systems theory of cellular reprogramming. Physical Review Research, 4(2), L022008.

Plugers, D., & Kaneko, K. (2025). Evolution of robust cell differentiation mechanisms under epigenetic feedback. arXiv preprint arXiv:2503.20651.

Ichikawa, K., & Kaneko, K. (2024). Bayesian inference is facilitated by modular neural networks with different time scales. PLOS Computational Biology, 20(3), e1011897.

11:00-​1​1:​40 野澤 恵理花 (山形大), Coupled Map Lattices for astronomical object formation and food processing: toward a consistent understanding and engineering application of complex phenomena

We introduce two models for astronomical object formation [1,2] and food processing [3] using coupled map lattice (CML), a powerful approach toward a consistent understanding of complex phenomena and its engineering application. (1) The CML for astronomical object formation simulates the chaotic dynamics and evolution of gas clumps without or containing a little dust with a minimal set of one Eulerian procedure for the flow formation of gas clumps due to gravitational interaction, and one Lagrangian procedure for the collision and mixture of gas clumps due to viscoelastic advection. Despite its simplicity, the CML successfully obtains four typical astronomical objects consistent with protoplanetary disk observations: a central star, Keplerian disk, spiral arms, and even stellar and substellar companions when gas clumps contain dust. All these formation processes are not based on the conventional disk gravitational instability but on the central star gravitational instability with high-dimensional chaotic gas ejection, namely the chaotic itinerancy. (2) The CML for food processing simulates the formation dynamics of diverse texture patterns that emerge spontaneously or self-organize during phase inversion processes going from fresh cream to butter via whipped cream under the physical whip processing with a minimal set of one Lagrangian procedure for the whipped cream formation due to whipping and two Eulerian procedures for the butter formation due to coalescence and flocculation. In the CML simulations, two well-known and different phase inversion processes are reproduced at high and low whipping temperatures. The overrun and viscosity changes simulated in these processes are consistent with those observed in experiments. By introducing a macroscopic state diagram, the viscosity-overrun plane, and a microscopic state diagram, the particle size-density plane, we characterize the two processes as the viscosity dominance and as the overrun dominance on the viscosity-overrun plane, and as the isodensity size dominance and as the isosize density dominance on the size-density plane, respectively, which succeeds in relating the macroscopic and microscopic state diagrams.

[1] E. Nozawa, "Coupled map lattice for the spiral pattern formation in astronomical objects", Physica D 405 (2020) 132377. https://doi.org/10.1016/j.physd.2020.132377

[2] E. Nozawa, "Jammed Keplerian gas leads to the formation and disappearance of spiral arms in a coupled map lattice for astronomical objects", Progress of Theoretical and Experimental Physics 2023 (6) (2023) 063A02. https://doi.org/10.1093/ptep/ptad064

[3] E. Nozawa and T. Deguchi, "Simulating phase inversion processes by coupled map lattice: Toward the theoretical design of food texture and quality in dairy processing from fresh cream to butter via whipped cream", The Journal of Chemical Physics 162 (6) (2025) 064902. https://doi.org/10.1063/5.0251375

11:45-12:25 柳田 達雄 (大阪電通大), Coupled-Map Modeling of Spatio-Temporal Pattern Formation

This talk presents a unified framework for spatio-temporal pattern formation using Coupled Map Lattices (CMLs). I demonstrate how simple local rules with diffusive coupling reproduce key phenomena such as phase separation, boiling, and cloud-pattern dynamics. I then introduce a preliminary typhoon model based on CML concepts, showing the emergence of self-organized vortex structures from minimal assumptions. Finally, I extend the method using the Coupled Map Gas (CMG) approach to simulate convection in spherical shells, enabling complex boundary conditions to be handled with simple, flexible dynamics. The approach highlights CMLs as powerful tools for linking microscopic rules to macroscopic patterns.

14:00-14:40 濱口 航介 (鹿児島大), Neural Dynamics Underlying Predictive Decision Making

We make decisions everyday. The brain dynamics make transitions from one state to another during the decision making, starting from sensory phase, action selection, motor preparation, and motor execution.   Here we investigate this neural dynamics of the 2ndary motor cortex (M2) of mice. Classically, M2 is thought to be involved in the motor planning and preparation, but accumulating evidence suggests that M2 also involves in the action selection itself. To select an action among many other options, one needs to know the value of each action. However, how and which neural circuit integrates these values and selects an action remains unknown. Combining reinforcement learning (RL) and two-photon calcium imaging, we found that the preparatory activity of neurons in M2, more specifically, the anterior-lateral motor (ALM) area initially encodes the value computed from the past history (retrospective value). After extensive training, the animal choice behavior becomes predictive. In this expert state, ALM neurons jointly encode the value computed from knowledge (prospective value). This preparatory activity in ALM emerges several seconds before the initiation of an action, and each neuron activity is selective to a specific type of action (e.g., left choice or right choice), thus ALM preparatory activity represents the future action and its value. Our findings showed that ALM is a critical neural hub that integrates retrospective and prospective values and biases the action toward the knowledge-dependent, predictive choice behavior (Hamaguchi et al., PNAS 2022). In this talk, I also introduce our recent analysis using the deep reinforcement learning (Deep-RL) to provide circuit level explanation of this predictive choice behavior. We show that Deep-RL spontaneously computes the value of action in the prospective manner. We will discuss the mechanism behind the emergence of prospective value coding in the mouse and artificial intelligence. 

1​4:45-15:25 北城 圭一(生理研), Metastable phase synchrony in the human brain

The human brain is a large-scale dynamical system that exhibits diverse nonlinear dynamics in its activity as measured by electroencephalography (EEG), including oscillations and synchronization. These dynamics play key roles in shaping the dynamical networks that underlie brain information processing. From a computational neuroscience perspective, I will introduce our research on the functional roles of metastable phase synchrony in EEG. Our empirical evidence suggests that metastable synchronization networks support flexible information processing, which in turn contributes to shaping individual brain functions and psychological traits. I will also present how these metastable dynamics relate to individual differences in the human mind and discuss their implications for the pathophysiology of psychiatric and neurological disorders.

1​5:40-16:20 山下 祐一 (国立精神・神経医療研究センター), Variational RNN-Based Digital Brain Modeling: Hierarchical Dynamics, State Estimation, and Virtual Intervention

This talk presents a “digital brain” approach based on a Predictive-Coding-Inspired Variational Recurrent Neural Network (PV-RNN), which reconstructs brain activity by estimating its underlying hierarchical and stochastic latent-state dynamics. The model is trained through variational free-energy minimization and self-organizes multi-timescale latent dynamical structures across three levels—local regions, functional networks, and global brain states. Using human stereo-EEG (SEEG), non-human primate electrocorticogram (ECoG), the model achieves high-precision reconstruction and prediction of neural activity. In addition, through data assimilation, it enables real-time inference of dynamically changing states of consciousness and behavior. Moreover, by manipulating the latent variables, the framework supports virtual interventions that emulate pharmacological and neuromodulatory manipulations, allowing simulation of functional changes at both the network and regional levels. This work advances our understanding of hierarchical organization, state transitions, and information-processing dynamics in the brain from a dynamical-systems perspective, while also highlighting the potential of such models for digital twin brains and the design of neural interventions.

16:25-17:05 栗川 知己 (はこだて未来大), Neural Dynamics Under Multi-Timescale Noise

神経系は極めて多様な時間スケールをもつシステムである。神経細胞の活動の時間スケールとシナプス変化の時間スケールだけでなく、神経細胞も場所により異なる時間スケールをもつことが知られている。一方でこの複数の時間スケールの協働がどのような機能を生み出しているかは、まだまだ明らかになっていない。

本発表では、まず現在取り組んでいる速いスケールとしての神経活動と遅いスケールとしてのシナプス変化にそれぞれノイズを考慮したシステムにおいて、連想記憶に着目することで、マルチタイムスケールのノイズが記憶の形成にどのように影響するかを示す。

また、時間があれば、神経活動に複数の時間スケールがある系で、working memory taskなどを行う際に、複数の時間スケールがどのような役割を果たすのか議論したい。

17:10-​1​7:50 鈴木 秀幸(阪大), 区分等長変換の弱カオス的ダイナミクス

区分等長変換（piecewise isometry）とはその名の通り区分的に長さを保つ写像であり、典型的な軌道はリアプノフ指数0となる。一方で、その不連続性により初期値鋭敏的な性質をあわせもち、「弱カオス」とも言うべき複雑なダイナミクスを示す。本講演では、1次元の区分等長変換において自己相似的な構造を見るとともに、多次元の例、工学システム等との関連や、離散拡散モデルへの応用を紹介する。

​​​

Monday, 8 December, 2025

1​​​0:00-​10​:​40 大平 徹 (名大), An exact solution of a non-autonomous delay differential equation and amplitude enhancement phenomena

Recently, one of the coauthors of this paper succeeded in obtaining an exact solution for a simple non-autonomous delay differential equation. Here, we report that when this delay differential equation is made to interact under appropriate parameter settings, an amplitude enhancement phenomenon occurs,with the amplitude increasing by as much as 108 compared to the non-interacting case. (This presentation is based on the following: K. Ohira, T. Ohira and H. Ohira,  "Amplitude Enhancements through rewiring of a non-autonomous delay system" Chaos 35, 041101 (Fast Track), 2025.  I can give my talk either in Japanese or English.)

1​​​0:45-​11​:​00 石井秀昌 (東大), Three distinct mechanisms of noise-induced escape in diffusively coupled bistable elements

When bistable systems are subject to dynamical noise, "noise-induced escape" from one state to the other is observed. We consider ensembles of noisy bistable elements with linear diffusive coupling. Despite the relatively simple setup, its analysis is a nontrivial task, as nonlinearity, coupling, and noise all play essential role in escape dynamics. We investigate dominant driving factors of escape processes to identify three distinct escape mechanisms.

11:00-11:15 吉原爽太 (名大), New Formulation of Dynamical Systems in One-on-One Pursuit and Evasion

J. C. Barton and C. J. Eliezer derived a set of simultaneous differential equations to formulate the one-on-one pursuit and evasion problem. At DD2024, we showed a dynamical system obtained from these equations. The system involves two new variables: the angular difference between the velocity vectors of the two players and the distance between them. This dynamical system explains the difference between circular and elliptical pursuit and evasion in terms of the existence of an equilibrium point. On the other hand, this system has a problem when the pursuer moves faster than the evader. In this case, the pursuer can catch up with the evader so that the inter-agent distance becomes zero. This leads to division-by-zero singularities and unphysical simulation behavior — for example, the pursuer may incorrectly begin to move away after contact. To overcome this difficulty, we reformulate the distance as an exponential function, where catching up corresponds to the exponent diverging. This idea also makes it possible to express the dynamics elegantly in the complex plane, where the magnitude and argument of a single complex variable represent the relative motion. Moreover, it can be approximated to a standard normal form using near-identity transformations, a method commonly used in the study of nonlinear oscillations.

11:15-11:30 小島瑛貴 (北大), Chaotic stochastic resonance in Mackey-Glass equations

Stochastic resonance (SR) is observed as switching dynamics between two quasi-stationary states in stochastic Mackey-Glass equations. We identify a new form of SR, chaotic SR, characterized by positive Lyapunov exponents, arising from the coexistence of SR and stochastic chaos. Unlike stable SR, which exhibits negative Lyapunov exponents, the resonance point for chaotic SR precedes the zero-crossing point of the largest Lyapunov exponent. Furthermore, we provide a theoretical estimate of the resonant periods for both stable and chaotic SR based on a linear mode analysis around an unstable fixed point. 

11:30-11:45 杉田　恭将 (一橋大), ロジスティック写像の反復関数系におけるいくつかの不変集合

反復関数系（Iterated Function System, IFS）はBarnsley(1988, Academic Press)によるフラクタル生成の典型例として知られるとともに、高次元力学系の簡約モデルとして理解することもできる。本研究では，異なるパラメータ値をもつ2つのロジスティック写像 $f$ と $g$ から構成されるIFSを分析する．一般にIFSは、写像$f$よりも複雑な振る舞いを示すが，我々はIFSの比較的単純な不変集合 $\Lambda$に着目する．そこで，写像$f$ と$g$それぞれの不変集合 $\Lambda_{f}$ と $\Lambda_{g}$ 間の交換構造に注目する．最も単純なケース1は，$\displaystyle \Lambda \setminus (\bigcup_{i}\Lambda_{f_{i}}\cup \bigcup_{j}\Lambda_{g_{j}}) = \varnothing$ となる場合である．Abbasi \& Gharaei \& Homburg(2018, Nonlinearity) は，IFSの不変集合が$f$の固定点と$g$の固定点のみにより構成される例を報告している．これはケース1の最も簡単な場合に相当している．ケース2は，$\displaystyle \Lambda \setminus (\bigcup_{i}\Lambda_{f_{i}}\cup \bigcup_{j}\Lambda_{g_{j}}) \neq \varnothing$ となる場合である．各ケースについて具体例を示す．

Tuesday, 9 December, 2025 

10:00-10:40 板尾 健司 (理研), Self-organized institutions in evolutionary dynamical-systems games

How do communities self-organize social institutions to manage a changing environment? Institutions are systems of shared norms and rules that regulate people's behaviors, often emerging without external enforcement. They provide criteria to distinguish cooperation from defection and establish rules to sustain cooperation. To address the mechanisms of their emergence, we introduce the evolutionary dynamical-systems game theory that couples game actions with environmental dynamics and explores the evolution of cognitive frameworks for decision-making. We analyze a minimal model of common-pool resource management, where resources grow naturally and are harvested. Players use decision-making functions to determine whether to harvest at each step, based on environmental and peer monitoring. After evolution, decision-making functions enable players to detect selfish harvesting and punish it by degrading the environment. This process leads to the self-organization of norms that define "cooperativeness" and rules of punishment, serving as institutions.

10:45-11:10 宮島 悠輔 (早稲田大), Mathematical models of amoeba-inspired combinatorial optimization machine

We propose mathematical models inspired by the survival strategies of slime molds, aiming to realize combinatorial optimization machines that combine high computational efficiency with low power consumption. The models are derived by applying three modifications that improve optimization performance to our previous model. Through this process, we can remove the strict constraint called the conservation law, which severely limits the device candidates for the physical implementation of the model. Furthermore, we perform a detailed analysis of these modifications to clarify why they improve the model’s optimization performance. These results are also expected to provide insight into the information processing of slime molds.

11:10-11:50 島田 尚 (東大), Dynamics of a Cross-feeding Ecosystem

In microbial community ecosystems, interspecies interactions through the leakage and uptake of metabolites are thought to be playing a dominant role. Model studies have been suggested that interactions through such leak-take interaction may contribute to the diversity and stability of the entire system (e.g. Yamagishi & Kaneko (2021), Clegg & Gross (2025)). In this talk, I will present our on-going study on the characteristic dynamics of a model ecosystem interacting with resource leakage and uptake. I will show the relationship between the dynamics of the system and the number of species that can coexist.

11:55-12:35 飯間 信 (広島大), Ascidian sperm navigation algorithm: spiral-sensing, fold-change detection, and canonical solution

Ascidian sperm employ a chemotactic strategy that integrates spiral swimming with intracellular $Ca^{2+}$ bursts to navigate complex chemoattractant fields.  We analyze this navigation algorithm using a minimal model and reveal that its underlying principle is Fold-Change Detection (FCD). FCD enables scale-invariant chemotaxis by responding to relative, rather than absolute, changes in stimulus concentration. The sperm spatially samples the concentration field along its spiral trajectory to infer the gradient direction. We propose an efficient canonical solution that describes the coupling of FCD-based signaling and simple locomotion, demonstrating a universal and robust strategy for navigation in fluctuating environments. This study is a collaboration with Prof. K. Shiba (U Tsukuba), K. Inaba (U Tsukuba), M. Yoshida (Kitasato U), and T. Nakagaki (Hokkaido U).

14:00-14:40 畝山 多加志 (名大), Coarse-Graining of Hamiltonian Dynamics into Langevin Equation with Fluctuating Diffusivity

A small tagged particle immersed in a liquid exhibits a random motion called the Brownian motion, although the whole system (the tagged particle and liquid molecules) obeys the deterministic Hamiltonian dynamics. The randomness arises from the coarse-graining which eliminates the degrees of freedom of liquid molecules and extracts the effective dynamics for the tagged particle. Following the standard procedure in nonequilibrium statistical mechanics, we can apply the projection operators to extract the dynamic equation for the tagged particle from that for the whole system. This procedure gives so-called the generalized Langevin equation (GLE), which contains the friction (damping) and random (fluctuating) terms with the memory effect. However, some systems cannot be successfully described as a simple GLE. For example, Brownian (or normal) yet non-Gaussian diffusion behavior is observed for some systems such as a particle in a supercooled liquid, but a simple GLE cannot reproduce the non-Gaussianity. The Langevin equation with fluctuating diffusivity (LEFD) is a useful model to describe such systems. We derive LEFD as a mesoscopic coarse-grained model of a microscopic Hamiltonian dynamics. We show that the projection operator method with auxiliary degrees of freedom gives LEFD (or a similar but slightly different dynamic equation model) instead of GLE.

14:45-15:25 山口 義幸 (京大), Universal discontinuous codimension-two bifurcation in long-range Hamiltonian systems

A long-range system may experience a bifurcation from the spatially homogeneous state to a clustered state.  The bifurcation is continuous in general, and it has been studied well. In this presentation we show emergence of a new bifurcation type by adding a condition at the critical point, and it is a codimension-two bifurcation. Interestingly, the added condition induces discontinuity of bifurcation. We also show that this codimension-two bifurcation is universal in an attractive system (mimicking a self-gravitating system), a repulsive system (plasma), and an Euler fluid. 

15:30-15:55 平岩 尚樹(JAXA), Probabilistic Invariant Sets Applied to Spacecraft Trajectory Design

TBA

16:00-16:50 辻井 正人(九大), Lyapunov exponent and bifurcation of chaotic attractors

We give an overview of recent progress in the study of partially hyperbolic dynamical systems and present the current state of the art in this field. We then discuss concrete families of dynamical systems that exhibit a chaotic attractor which persists stably under perturbations, while its (second) Lyapunov exponent changes its sign without altering the qualitative nature of the attractor.

17:00-17:50 首藤 啓(東京都立大学), Eigenstate flooding and amphibious complex trajectories 

TBA

Wednesday, 10 December, 2025 

1​​​0:00-​10​:40 佐藤 譲 (北大), Intermittency and supertransients in dynamical systems: a review

We review representative forms of intermittent and transient dynamics that have been studied in nonlinear physics, presenting concrete examples. In particular, we introduce the theoretical background and phenomenology of Pomeau–Manneville intermittency, on–off intermittency, in–out intermittency, crisis-induced intermittency, supertransient chaos, spatiotemporal intermittency, and chaotic itinerancy. This talk is based on a recent invited review article published in the Journal of the Physical Society of Japan.

1​​0:45-​11​:​25 末谷 大道 (大分大), Impact of weak generalized synchronization on time series forecasting using reservoir computers

We investigate the relationship between chaotic time series forecasting performance and the dynamical properties ofecho state networks (ESNs) from the viewpoint of generalized synchronization (GS). By treating the ESN as a response system driven by chaotic input, we analyze the existence and quality of GS using conditional Lyapunov exponents (CLEs) and a replica synchronization error-based detection method. We distinguish between strong GS, where the reservoir state depends smoothly on the input, and weak GS, where the synchronization function loses its smoothness and becomes sensitive to perturbations. Our results show that forecasting accuracy does not peak at the edge of conditional stability, but instead near the transition from strong to weak GS. Additionally, overly smooth synchronization in the strong GS regime can reduce input sensitivity and degrade performance. These results support the value of viewing reservoir computing as a framework for understanding the role of nonlinear dynamics in information processing. Rather than focusing solely on machine learning performance, our results highlight that a balance between dynamical richness and input sensitivity enables flexible information processing in non-autonomous dynamical systems, such as neural systems.

11:30-12:10 秦 浩起 (鹿児島大), 高次元データによる低次元カオスの予測

カオス力学系の時間発展について，高次元データと予測という視点から少し考察する。散逸系カオスでは，その時間発展は（相対的に）低次元の力学系に従う（例，ローレンツカオス，B-Zカオス，水滴系カオス）。しかし，その低次元力学系を見出すことは一般に容易ではない。ここでは，<低次元力学系から生成された>高次元データから時間発展を予測すること，すなわち，低次元力学系を近似的に見出すことについて，数値実験を基にした予備的結果を紹介したい。

Home

Supported by

• JSPS Grant in aid for Fund for the Promotion of Joint International Research (Fostering Joint International Research (B)) 22KK0159 (PI: Hiromichi Suetani)

• JSPS Grant in aid for scientific research (B) 25K03197 (PI: Hiromichi Suetani)

• JSPS Grant in aid for scientific research (B) 21H01002 (PI: Yuzuru Sato)

• Faculty of Science, Kagoshima University

• Research Institute for Electronic Science, Hokkaido University